The number is 3.4. Across 338,231 older motorcyclist casualties recorded in Taiwan between 2011 and 2022, the share of casualties who died climbed with every age band: 0.5% at 65–69, then 0.8%, 1.1%, 1.3%, and 1.7% at 85 and over. From the youngest band to the oldest, the chance that a crash was fatal rose 3.4 times. Not because the crashes got worse. Because the bodies did.
That is the curve this piece is about, and it is the one the rental shop will never draw on the form. The crash you survive at 60 is the crash that kills you at 80. Same scooter, same intersection, same impact — a different outcome, scored by the only variable that moves: the age of the person the impact lands on.
The curve
The source is a 2025 PLOS One study of older motorcyclists in Taiwan, drawn from the National Traffic Crash Dataset held by the National Police Agency. It is not a survey or a sample. It is the police record of 338,231 casualties aged 65 and over across twelve years: 2,703 of them fatal, 335,528 not. Of the whole population, 11,479 (about one in thirty) were aged 85 or over, and that oldest band carried the highest fatality rate of all.
Read the case-fatality column on its own and the shape is unmistakable. It rises every single step.
Source: Wang et al., PLOS One 2025 — old motorcyclists in Taiwan, 2011–2022 · checked 2026-06-05
A raw case-fatality rate confounds age with everything age drags along: slower riding, shorter trips, fewer night rides, more careful behaviour. So the authors ran the regression. Holding helmet use, licensing, sobriety, and the other recorded covariates constant, they read off the odds of a fatal injury for each age band against a 65–69 reference. The gradient survived the adjustment intact. At 70–74 the odds ran 1.30 times the reference; at 75–79, 1.70; at 80–84, 1.96; at 85 and over, 2.43 (95% CI 2.051–2.879). The oldest riders were not dying more because they rode worse. They were dying more because the same crash found a frailer body.
Source: Wang et al., PLOS One 2025, Table 2 · checked 2026-06-05
Two readings of the same data, the raw rate and the adjusted odds, point the same way. That is what a real gradient looks like. The behaviour washes out and the age stays.
Why the same crash kills later
The paper does not leave the mechanism to inference. Older motorcyclists are, in the authors’ words, “physiologically more vulnerable to sustain severe injuries owing to factors such as decreased bone density and slower healing”, carrying decreased physical resilience and a higher prevalence of comorbidities. The crash does not change. The threshold at which the crash becomes fatal does.
Think about what a 35 km/h low-side does to a body. At 60, it is a broken collarbone, a road rash, six weeks and a bill. At 82, the same fall is a fractured hip, and the fractured hip is its own actuarial event: pooled one-year mortality after a hip fracture across the Asia-Pacific runs 17%, from 10.8% in Singapore to 23.8% in New Zealand. Lower bone density, slower healing, less reserve. That is the paper’s own account of why a constant amount of crash energy stops being survivable, and it is why the fall is scored differently at 82 than at 60. None of that is bad luck. It is the predictable arithmetic of a body with less margin absorbing a constant amount of energy.
This is why the curve only points one way. There is no age band where survivability recovers, no decade where the body gets tougher. The line in the chart has no local minimum after 65 because the underlying physiology has none. A relocation plan that assumes the ride stays the same for twenty years is assuming away the one thing that is guaranteed to change. The alternative outcome is not survival unchanged but survival without the ride — a loss with its own measured cascade, landing in the same decade.
The Taiwan caveat
Be exact about what the figures are. The 338,231 casualties are Taiwanese, and so are the case-fatality rates, the 3.4× rise and the 2.43 at 85+. How many published studies measure the same curve for retired foreigners riding scooters in Southeast Asia? Zero. Attaching a precise expat number to this one would be manufacture, not measurement.
What licenses the transfer is the mechanism. Decreased bone density, slower healing and less physiological reserve do not stop at a border, and Taiwan has a denser trauma-centre network, higher helmet compliance and better-engineered roads than most of the corridors a relocated foreigner actually rides. So the gradient is a floor. Whatever the true SE Asian curve is, it sits above this one.
Where the curve gets ridden
The curve is being ridden where the road already does most of its killing on two wheels. Motorcyclists are 83.8% of everyone killed on Thai roads, and the Philippine Department of Health reports 65% of road-crash victims are motorcycle riders. That share follows the fleet, in a region where about 87% of Thai households owned a motorcycle in 2023 and nearly half of Filipino households ride one, and where losing the ability to ride at all is its own dated event. The relocation quietly assumes you keep riding for life, and the age gradient is what that assumption costs.
The population baselines under those shares are not gentle ones. The WHO Global Status Report on Road Safety 2023 (data year 2021) records Thailand at 25.4 road deaths per 100,000, among the highest in the world; Thai national reporting counted 14,144 motorcyclist deaths in 2024 out of 17,447 road deaths in total. The Philippines records 9.7 per 100,000. These are the baselines this site has costed elsewhere as the foreigner road-death multiplier. That piece asks how much more likely a foreign pensioner is to crash at all. This one holds the crash constant and asks how much more likely it is to be the last one.
What stacks on top of age
Age is not the only coefficient in the Taiwan table, and the others compound it rather than replace it. In the same model, the adjusted odds of a fatal injury ran 3.58 for not wearing a helmet, 3.86 for drunk riding, and 1.38 for riding unlicensed, each independent of the age term, each landing on top of it.
| Factor | AOR (fatal injury) | What it means on the road |
|---|---|---|
| Age 85+ (vs 65–69) | AOR (fatal injury) 2.43 | What it means on the road The oldest band against the youngest, behaviour held constant. |
| Age 80–84 (vs 65–69) | AOR (fatal injury) 1.96 | What it means on the road Already nearly doubled before the eighties begin. |
| Not wearing a helmet | AOR (fatal injury) 3.58 | What it means on the road In Thailand, 84% of hospitalised riders were unhelmeted at the crash. |
| Drunk riding | AOR (fatal injury) 3.86 | What it means on the road The single largest coefficient in the table. |
| Unlicensed riding | AOR (fatal injury) 1.38 | What it means on the road The default state of most foreigners on rental scooters. |
Source: Wang et al., PLOS One 2025, Table 2; Thai DDC helmet figure 2020–2024 · checked 2026-06-05
The combinations are the part to sit with. An 85-year-old riding unhelmeted is multiplying two of the largest numbers in the table against each other. And the helmet coefficient is not hypothetical in the destination: 84% of motorcyclists hospitalised in Thailand from 2020–2024 were not wearing one. Foreign rental-scooter riders are not a safer subset of that population; the rental shop hands over the key and rarely the helmet, and never checks the licence. The base curve already slopes up with every year. The behaviour the destination normalises stacks on top.
There is a useful split hidden in that table, and it is the closest thing to good news the data offers. Two of the three behavioural coefficients are modifiable. A rider can wear the helmet, which collapses the 3.58. A rider can stay sober on the bike, which collapses the 3.86. A rider can hold a proper licence, which removes the 1.38 and, not incidentally, the insurance void that sits behind it. Those are real levers, and they are large.
The age term is not a lever. It is 1.30, then 1.70, then 1.96, then 2.43, and no behaviour moves it, because the thing it measures is the body’s falling capacity to survive a given injury. This is the part worth being honest about. The advice the rental shops and the brochures will give (wear a helmet, do not drink and ride, get a licence) is correct, and it addresses the coefficients you can change. It is silent on the one you cannot, and the one you cannot is the one that compounds with every year you stay. You can ride the curve more carefully. You cannot ride a different curve.
The curve over a relocation
Put the gradient on the timeline the relocation actually runs on, and the cost shows up as a slope rather than a number. A common pattern: the move happens around 60, the scooter is the daily vehicle from day one, and the plan assumes it stays that way. Hold the crash type constant and walk the same rider up the age bands the Taiwan curve measured.
| Decade on the scooter | Case-fatality per crash | vs the first years | What has changed |
|---|---|---|---|
| Early sixties (65–69 band) | Case-fatality per crash ~0.5% | vs the first years 1.0× | What has changed The years the relocation maths actually prices. |
| Early seventies (70–74) | Case-fatality per crash ~0.8% | vs the first years 1.6× | What has changed The same scooter, a body with less reserve. |
| Late seventies (75–79) | Case-fatality per crash ~1.1% | vs the first years 2.2× | What has changed Bone density down, healing slower, comorbidities accruing. |
| Early eighties (80–84) | Case-fatality per crash ~1.3% | vs the first years 2.6× | What has changed A survivable fall becomes a hip fracture and its cascade. |
| Eighty-five and over | Case-fatality per crash ~1.7% | vs the first years 3.4× | What has changed The crash the early-sixties rider walked away from. |
Source: Composed from the Taiwan PLOS One 2025 case-fatality gradient; gradient transfer, not a measured expat figure · checked 2026-06-05
The right-hand column is the honest read. Nothing in the relocation changes the road or the scooter. What changes is the rider, every year, in one direction. The plan that pencils in lifelong two-wheel mobility has priced the 0.5% years and quietly carried the 1.7% years at the same rate. The risk did not stay flat. It bent upward the whole time, and the brochure photographed it at the bottom of the curve.
Contest the gradient
The transfer lives or dies on a few honest assumptions. State them, and the direction each moves the conclusion.
- The population is Taiwanese. The case-fatality rates and the AOR are measured on Taiwan, 2011–2022, not on SE-Asia expats. Direction: no expat-specific curve exists to confirm the level.
- Case-fatality ≠ crash probability. This curve is conditional on a crash having happened. It says nothing about how often the crash occurs. The chance of crashing at all is the separate foreigner road-death multiplier; the two compound. Direction: the combined risk is steeper than either factor alone.
- Confounding washes out, but not perfectly. The adjusted odds (2.43 at 85+) control for helmet, licence, sobriety, and recorded covariates, but not for comorbidity load or crash partner, which the authors flag as missing. Direction: unmeasured frailty would strengthen the age term if it did anything at all.
- The relocation transfer is illustrative. Walking one rider up the Taiwan bands assumes the same crash mix at every age. Real older riders self-select toward shorter, slower, daytime trips, which lowers exposure. Direction: fewer crashes, each one just as likely to kill. The curve is unmoved.
- Behavioural coefficients stack onto the age curve; they do not replace it. The unhelmeted (3.58) and drunk (3.86) AORs describe a population, and no individual rider is a population. Direction: they raise the ceiling for the rider who carries them and leave the floor, the age curve itself, in place.
What the curve argues
It does not argue that you should not ride. It argues that the ride is a depreciating contract, and what depreciates is the rider while the bike stays exactly what it was. The shop hands the same key to the same 125cc scooter at 62 and at 82. The road is more or less the same road. The thing that is not the same is the body that takes the impact when the impact comes, and the data is clear that the body gets less survivable every year, monotonically, with no decade off.
The relocation maths that prices the early years and assumes they repeat has bought the cheap end of the curve and ignored the expensive end. The expensive end arrives on schedule, in the place where the trauma centre is further away, the helmet less likely worn, and the family a long-haul flight from the bedside. It also arrives in the decade where private cover thins out, because the insurers have walked off the back of the same age curve — the rider who most needs the policy is the one least likely to still hold it.
It is arithmetic, then, and not a warning. The curve is real. It only points one way. Decide whether the rider twenty years from now, the one who takes the fall you would shrug off today, is a person you have actually planned for — or a person the brochure cropped out of the frame.